*POWER CALCULATIONS



*Based on Gangl et al because they are thematically closest to what we are doing, and they also report sd:s.


*Unclear whether nationalism scale was normalized or not, so the power calculations are based on national pride and identification.

				National Fictitious
				Flag	 Flag
				M (SD)		M (SD)
National pride	5.32 (1.45)	4.48 (2.10)  /// 5.32-4.48=0.84/6=0.14=14% difference
Identification 	5.02 (1.93)	4.63 (1.59) ///5.02-4.63=0.39/6=0.065=6.5% difference


*National flag --> National pride: 
power twomeans 4.48 5.32, power(0.8 0.9) sd(1.7) graph
*sample size of 130 to 180

power twomeans 5.02 4.63, power(0.8 0.9) sd(1.7) graph
*sample size of 600-800 


*National landscape --> Nationalism: 

	National Landscape	Foreign Landscape
				M (SD)		M (SD)
National pride	4.81 (2.03)	4.66 (1.58) /// 4.81-4.66=0.15/6=0.025=2.5%
Identification  4.92 (1.73)	4.03 (2.01) /// 4.92-4.03=0.89/6=0.015=15%

power twomeans 4.81 4.66, power(0.8 0.9) sd(1.7) graph
*sample size of 4000-5000

power twomeans 4.92 4.03, power(0.8 0.9) sd(1.8) graph
*sample size of 130-170


*******************************

*Based on Butz et al. 2007:


			Flag   No flag
			M 	SD 		M	SD
Ptriotism	4.09 0.56 4.22 0.57 /// 4.22-4.09=0.13/4=0.0325=3.25%
Nationalism	2.79 0.74 2.60 0.67 /// 2.7-2.60=0.19/4=0.0475=4.75%
Anger etc	2.77 1.62 3.55 2.13 /// 3.55-2.77=0.78/9=0.087=8.7%

power twomeans 4.09 4.22, power(0.8 0.9) sd(0.56) graph
*sample size of 600-800

power twomeans 2.79 2.60, power(0.8 0.9) sd(0.7) graph
*sample size of 430-580

power twomeans 2.77 3.55, power(0.8 0.9) sd(2) graph
*sample size of 210-280



*Results reveal that to detect effects of similar magnitud as the largest effects in these experiments, ///
i.e. around 14-15%, a sample of less than 200 respondents would suffice. 

*To detect medium sized effects according to these studies, i.e. effect around 3-6%, ///
a sample size of around 500-600 respondents would suffice. 

*To detect small effect sizes, i.e. smaller than 3%, a sample of no less than 1000 respondents is needed.

*But all this depends on standard deviations. 
